Secondary outcomes

Reductions in energy content per individual swap offer were analysed using multilevel linear regression to assess (i) whether offering swaps significantly reduced the energy content of participants’ purchases (original choices versus final purchases), (ii) whether any reduction was moderated by swap frame and (iii) what the potential impact of the offering swaps would have been, were all swaps accepted (original choices versus lower-energy alternatives, regardless of whether they were accepted or not) and (iv) what the potential additional impact of offering swaps would have been, had all swaps been accepted (final purchases versus lower-energy alternatives). As before, swaps were nested within participants, and models controlled for the number of lower-energy alternatives offered per swap, and the number of swaps offered to each participant. Time was accounted for as a dummy variable (e.g., original choices vs. final purchases). As two analyses were conducted for each of the outcomes (observed energy change—analyses i and ii; potential energy change—analyses iii and iv), the critical significance level was adjusted using a Bonferroni correction to .025.

Reductions in energy content per total shopping basket were analysed using t-tests to assess the impact of offering swaps on total purchased energy, and the potential impact of offering swaps had all swaps been accepted. Change in energy content of shopping baskets from pre- to post-swap offer was assessed, and linear regression was used to see whether this differed by swap frame group. An additional model explored whether demographic variables (age, gender, BMI, education level, income category) influenced basket energy change. As three analyses were performed on observed change, the critical p value was adjusted using a Bonferroni correction to .017. Finally, t-tests were used to assess the effect of offering swaps on total basket saturated fat, sugar and salt (all in grams). As three comparisons were conducted on these nutrients, the critical significance value was adjusted using a Bonferroni correction to .017.